Want to read Slashdot from your mobile device? Point it at m.slashdot.org and keep reading!


Forgot your password?

Comment Re: Reputation (Score 1) 131

I said *quickly* gauge the paper. Of course I can read it, but there's a lot of papers and my time is limited. If it's published in a reputable venue, that's an endorsement that helps me order the papers preliminarily.

Of course, there are other sorting criteria like a number of citations, but they have their own issues - time lag, variability across (sub)fields, biases towards certain kinds of papers, etc.

Comment Reputation (Score 1) 131

Of course, people are trying to explore other options - e.g. http://www.iclr.cc/doku.php?id...

The problem is reputation. *Where* was the paper published carries huge weight on both the repute of the paper and change in repute of the author, because noone figured out better ways to quickly judge a result than by the venue (which implies certain acceptance rate and level of peer review standards). If you move from the established institutions to elsewhere, you need to build up your repute from scratch and until you do...

Well, it just takes long time. That means decades when it's not a new emerging field. Many decades when the academics are particularly conservative.

Comment Re:So... (Score 1) 57

A is right. The task is identifying pedestrians. Miss rate means that the algorithm fails to identify the pedestrian. Lower number is better.

Also, when the algorithm fails, it doesn't mean the car will just happily drive through the pedestrian!

First, most pedestrians are not at the road, and the ones at the road should be actually easier to identify as they won't blend that much (I guess). Second, there are probably already many components that already identify obstacles and try to avoid them. Identifying pedestrians specifically is helpful to predict their movement, choosing another obstacle if you are going to hit some obstacle anyway, etc.

Comment Re:So... (Score 3, Insightful) 57

No. As usual, the summary is confusing as it gives numbers for the *older* methods, but the current Google's method is: "The resulting approach achieves a 26.2% average miss rate on the Caltech Pedestrian detection benchmark, which is competitive with the very best reported results. "

So, there's 26.2% chance that on a single particular image, you miss the pedestrian (at the same time, it seems that in about 15-20% images it sees a pedestrian that is in fact a shrubbery or whatever). This is an academic dataset, and in reality you will have a video feed. AFAICS it's not clear how the precision translates when you have a sequence of many pictures of the pedestrian - whether you will have much higher chances to spot them at least on some of them, or if it's more of a systematic problem and khaki-clothed people just don't stand a chance.

Comment What they actually did. (Score 4, Informative) 66

The paper is a bit confusing at first, and the /. summary doesn't help. Basically, they developed a sorting criteria to reduce the amount of work for the editors. In an isolated comparison of two jokes, the funnier joke wins 64% of them on average; this is quite better than a coin!

To get a sorted list, they run a "comparison tournament" between the jokes. The 55.8% number means that the funniest joke is in the top 55.8% of the list on average; if we are willing to occasionally miss a brilliant joke, we can cut the list in a little more than half and still keep most of the great jokes.

The full paper is http://research.microsoft.com/...

Comment qmail and Microsoft (Score 5, Informative) 85

Yeah, you seem to want to have your cake and eat it too. Doesn't produce a lot of sympathy. Think again about how to make your software free but still want users to pay. What about keeping value-adding plugins or frontends closed and opening the core? If you open source but limit ability of people to make use of the core, what exactly do you expect to gain from such a "community"?

Still, take a look at the licence of qmail. This worked not so bad for them, and might be the right equilibrium. If you just want legalese for your scenario, take a look at Microsoft's Shared Source licences.

Comment Re:"No idea how... the brain works" (Score 4, Informative) 230

(I work in this area of research.) You are right, the paper is about just a sequence-to-sequence transformation model that learns good replies for inputs but is not actually "understanding" what is going on.

At the same time, we *are* making some headways in the "understanding" part as well, just not in this particular paper. Basically, we have ways to convert individual words to many-dimensional numerical vectors whose mathematical relations closely correspond to semantics of the words, and we are now working on building neural networks that build up such vectors even for larger pieces of text and use them for more advanced things. If anyone is interested, look up word2vec, "distributed representations" or "word embeddings" (or "compositional embeddings").

If you already know what word2vec is, take a look at http://emnlp2014.org/tutorials...

Comment Java API: Copyrighted, but hope for fair use! (Score 4, Informative) 223

TL;DR: US executive shares the appeals court opinion that APIs are copyrightable, but that does not mean the copyright is enforceable - there will be another court case that will be about if it's fair use to re-implement the (copyrighted) API.

Here is maybe the most important paragraph (italics mine):

Despite the inherently functional character of all computer code, the Copyright Act makes clear that such code can be copyrightable. Nothing about the declaring code (API declarations) at issue here materially distinguishes it from other computer code ... . Although petitioner has raised important concerns about the effects that enforcing respondent's copyright could have on software development, those concerns are better addressed through petitioner's fair-use defense, which will be considered on remand.

The brief is quite well readable (modulo the awful scribus ui), try it!

Comment Re:bye (Score 2) 531

I do use the memory, thank you very much. I just use the computer for something else than web browsing too. I do realize it's getting uncommon (and don't actually even get *that* grumpy about it, just have my different set of preferences).

When you see a guy in sibling comment complaining "In addition, I hate the extremely long time for startup and new tab creation, which is accompanied by constant disk grinding." - well, that's exactly the memory problem, which now translates to bad user experience. As you suggested: The OS swapped it out as another app needed it.

Comment Re:bye (Score 2) 531

Oh and also the fact that middle click inside website does not load URL from clipboard. It works on a favicon - except in case of verified identity SSL servers, there's no favicon.

About the tabs, frankly, for me (but clearly not just me) a more flexible paradigm which blends seamlessly the concept of tabs and bookmarks (and ideally full-text search over my "bookmarked tabs") would be awesome. I'm a pack-rat and would like to archive whole tab trees for later, see them among the other pages, but not take memory+CPU now. I think there's an actual and large market gap here.

Machines have less problems. I'd like to be a machine. -- Andy Warhol